You add the posit marketplace, install r-lib, and invoke the skill at an opportune moment (or claude code runs it for you with the right trigger words).
The README has some instructions.
github.com/posit-dev/sk...
I should add the "Kratt momentum" page is from 2018.
and
"An interesting aspect of the kratt is that it was necessary for it to constantly keep working, otherwise it would turn dangerous to its owner."
Estonians were truly ahead of the #AI curve: "in Estonia [the Kratt, a golemesque creature] is used as a metaphor for AI and its complexities. For instance, algorithmic-liability law is also called the Kratt law."
en.wikipedia.org/wiki/Kratt
use this skill for some extra help on the arcane: github.com/posit-dev/sk...
Bullshit Benchmark tests whether models challenge nonsensical prompts instead of confidently answering them.
The #AI benchmark that matters
github.com/petergpt/bul...
Evergreen: bsky.app/profile/ulri...
Vibecoding in #Rstats visualized
Zalgo-style glitch text: \glitch{UNFINISHED}
New \glitch{} #latex command just dropped to freak out your coauthors. Tells exactly how the paper is going.
gist.github.com/statzhero/ea...
Last week we released dplyr 1.2.0, but we left off something VERY important 🙂
`dplyr::if_else()` and `dplyr::case_when()` are now up to 30x faster and use 10x less memory!
We dive into how we achieved these numbers in this new #rstats post!
tidyverse.org/blog/2026/02...
Github error 500
is github down?
Child at the podium: “A woo woo woo.”
Mamdani: That’s how I felt when we came up with this plan. Together, we will expand the idea of what is possible in our city—and what sounds and noises we can make at a press conference.
It's also titled: "Olympic ski jumpers to be investigated for ‘injecting penises with acid’"
Tagging you here as we were talking about the news: @paulgp.com
The usual #AI pros and cons apply also for Claude Finance (as of today): it is remarkably fast and capable, but still requires supervision and will make mistakes. The spreadsheets look spiffy though.
statzhero.github.io/ulrichatz/blog/opus-finance.html
Today, Anthropic released Claude Opus 4.6 with a #Finance plugin. I tested it on capitalizing VW's R&D expenses. In one vague prompt I got a full spreadsheet with amortization schedules and ROIC calculations.
Happening right now, I could live post it.
It's also confusing to me what agents are doing at any moment. Not unlike a classroom.
Actually, the one thing I will try it for is TikZ graphics because Claude needs an agent team member with “eyes”.
I saw this and was like: I am become RA manager, destroyer of tokens.
but I’m happy with my subagents for now.
What’s the use case?
Is there an #rstats equivalent?
A correct take
`seed(42)` is the new em dash.
One more q: will it work without the Letta max plan? 💀
Prompt injection I guess would be top of mind. Or it overwrites the permission settings.
One of the greatest tools for scientific production relaunched: #Zotero 8.
www.zotero.org/blog/zotero-8/
Looks awesome. Can you provide a high-level comment on its security, ie if it is beyond the usual dangers of using Claude Code?
🧪 Validation RMCP has been tested with real-world scenarios achieving 100% success rate: ✅ Business Analysts: Sales forecasting with 97.9% R², $4.70 ROI per marketing dollar ✅ Economists: Macroeconomic analysis confirming Okun's Law (r=-0.944) ✅ Data Scientists: Customer churn prediction with 100% accuracy ✅ Researchers: Treatment effect analysis with significant results (p<0.001)
Sure, buddy.
#rstats #mcp
The design aims to follow the approach in Ian Gow and Tony Ding's "Empirical Research in Accounting: Tools and Methods" book, which emphasizes reproducible workflows.
It provides:
- Secure credential management via your system keyring
- Pre-built functions for Compustat, CCM, and linking tables
- Standard default research filters, and an argument to fill in missing SIC fields
